Plaque-targeted, proteolysis-resistant, activatable and MRI-visible nano-GLP-1 receptor agonist targets smooth muscle cell differentiation in atherosclerosis

Background: Glucagon-like peptide-1 receptor (GLP-1R) agonists are powerful glycemia-lowering agents, which have systematically been shown to lower cardiovascular events and mortality. These beneficial effects were difficult to pinpoint within atherosclerotic plaque due to lack of particular specificity of such agonists to the vascular cells and an inadequate understanding of the GLP-1R expression in atherosclerosis. Here, we hypothesized that the direct engagement of the GLP-1R in atherosclerosis by targeted agonists will alleviate vascular inflammation and plaque burden, even at a very low dose. Methods: The expression of GLP-1 receptor (GLP-1R, Glp1r mRNA) in human lesions with pathologic intimal thickening, Apoe-/- mouse atheroma and cultured immune/non-immune cells was investigated using genetic lineage tracing, Southern blotting and validated antisera against human GLP-1R. Protease-resistant and “activatable” nanoparticles (NPs) carrying GLP-1R agonist liraglutide (GlpNP) were engineered and synthesized. Inclusion of gadolinium chelates into GlpNP allowed for imaging by MRI. Atherosclerotic Apoe-/- mice were treated intravenously with a single dose (30 µg/kg of liraglutide) or chronically (1 µg/kg, 6 weeks, 2x/week) with GlpNP, liraglutide or control NPs, followed by assessment of metabolic parameters, atheroma burden, inflammation and vascular function. Results: Humal plaque specimens expressed high levels of GLP-1R within the locus of de-differentiated smooth muscle cells that also expressed myeloid marker CD68. However, innate immune cells under a variety of conditions expressed very low levels of Glp1r, as seen in lineage tracing and Southern blotting experiments examining full-length open reading frame mRNA transcripts. Importantly, de-differentiated vascular smooth muscle cells demonstrated significant Glp1r expression levels, suggesting that these could represent the cells with predominant Glp1r-positivity in atherosclerosis. GlpNP resisted proteolysis and demonstrated biological activity including in vivo glycemia lowering at 30 µg/kg and in vitro cholesterol efflux. Activatable properties of GlpNP were confirmed in vitro by imaging cytometry and in vivo using whole organ imaging. GlpNP targeted CD11b+/CD11c+ cells in circulation and smooth muscle cells in aortic plaque in Apoe-/- mice when assessed by MRI and fluorescence imaging. At a very low dose of 1 µg/kg, previously known to have little effect on glycemia and weight loss, GlpNP delivered i.v. for six weeks reduced triglyceride-rich lipoproteins in plasma, plaque burden and plaque cholesterol without significant effects on weight, glycemia and plasma cholesterol levels. Conclusions: GlpNP improves atherosclerosis at weight-neutral doses as low as 1 µg/kg with the effects independent from the pancreas or the central nervous system. Our study underlines the importance of direct actions of GLP-1 analogs on atherosclerosis, involving cholesterol efflux and inflammation. Our findings are the first to suggest the therapeutic modulation of vascular targets by GlpNP, especially in the context of smooth muscle cell inflammation.

. Glp1r mRNA expression in mouse and human cells and tissues A) The expression of Glp1r mRNA transcripts in mouse cultured primary immune cells was null or very low as shown by quantitative real-time PCR amplification curves. The cutoff was set at the 30th cycle (arrowhead and dotted line) for reliable analysis. Atherosclerosis-relevant conditions were tested: BMDM, P-Mac, BMDDC, BMMo -bone marrow derived macrophages, peritoneal macrophages, bone marrow derived dendritic cells and monocytes respectively, were cultured at high glucose conditions or additionally treated with 0.5 µg/mL oxidized LDL (oxLDL), 10 4 /10 6 cells UV-induced apoptotic bodies (UV-AB), 100 ng/mL LPS plus 100 U/mL interferon gamma (LPS/IFNγ), or 10 ng/mL interleukin 4 (IL-4). Whole pancreas mRNA served as a Glp1r positive control. B) Splenocytes from mice expressing Cre recombinase under the Glp1r promoter crossed with a ROSA26 eYFP (Glp1r-Cre/R26R-EYFP) were analyzed by flow cytometry co-staining with antibodies against CD45 (pan myeloid cells), CD19 (B cells), CD3 (T cells) and CD11b (macrophages). Gating was selected as indicated. Glp1r + cells were detected in gated populations through intrinsic eYFP signal. C) Glp1r mRNA expression in mouse whole atherosclerotic aortas is comparable to those in organs known to express Glp1r. Relative mRNA expression is calculated vs. control ΔCt values obtained from the whole aorta of C57BL/6 mouse. D) The expression of various housekeeping genes among 30 atherosclerotic tissues was analyzed by   . Isotype control staining in human coronary artery. Immunofluorescent staining images of αSMA (green) and mouse IgG1 (red, isotype for GLP1R) are shown. Nuclei were counterstained with DAPI (blue).

Figure S5. Specific cholesterol loading conditions induce mouse aortic smooth muscle cells (MASMC) de-differentiation. A)
Various cholesterol loading conditions were tested to overload MASMC with cholesterol as indicated. Modified low density lipoproteins (LDL), including aggregated, acetylated (Acetyl-), loaded with 7-ketocholesterol (7KC), oxidized (oxLDL) and very low density (vLDL) were incubated with MASMC at 100 µg/mL for 96 h. This was compared with MASMC incubated with free soluble cholesterol (cyclodextrin complex, Sigma C4951) at 10 µg/mL or cholesterol crystals at 200 µg/mL. Intracellular cholesterol deposition was analyzed by BODIPY staining (green) after cellular fixation. Nuclei were counterstained with Hoescht (blue). B) The MASMC phenotype was confirmed by RNA sequencing (RNAseq) after loading with C4951 for 96 h. 198 differentially expressed genes were found, including downregulated SMC markers and upregulated macrophage (Mac) markers as indicated in bar graphs. Major cholesterol biosynthesis genes were downregulated. Gene ontology pathway analysis indicated various cholesterol biosynthesis pathways significantly downregulated, however lipid transport, autophagy and inflammation upregulated. Averages from two independent experiments (two untreated and two C4951-treated) are shown with an error bar indicating the spread of the data.  Supplementary Table S4. B) Heatmap representing the expression of smooth muscle and endothelial cell markers in the endothelial cell line 3B11 and primary carotid artery endothelial cells treated as indicated and analyzed by quantitative PCR (left). Statistically significant differences between groups are indicated by squares and circles and additionally plotted as a bar graph (right). NE -not expressed. C) Immunofluorescence staining in primary mouse aortic smooth muscle cells (MASMCs) using antibodies against a marker of adult smooth muscle SM22 (transgelin, red) and co-staining with anti-mouse GLP-1R mAb 7F38 (green). Nuclei were counterstained with DAPI (blue).

Figure S7. Immunofluorescence analysis of GLP-1R expression in mouse atherosclerosis.
Immunofluorescence staining was performed on brachiocephalic artery (BCA) sections from Apoe -/mice high-fat fed for 30 weeks using anti-mouse monoclonal GLP-1R antibody mAb 7F38 (red). The sections were co-stained with rabbit anti-CD68 (macrophage marker, green) displayed at the top set of images, or with rat anti-VE-cadherin (endothelial cell marker, green) at the bottom set of images, as indicated. Corresponding Pearson's correlation coefficients of colocalization are shown at the right. Metric matrices for two reporter channels were calculated using Pearson correlation coefficient (PCC) and show percentile (%) of pixels for signal intensity in a given image. PCC=1 means complete co-localization and PCC=-1 defines complete lack of co-localization. Figure S8. A) Proposed mechanism of action of activatable GlpNP. GlpNP has three layers: i) latex (polystyrene) core with fluorochrome (blue), ii) gelatin (orange), iii) polyethylene glycol (black), iv) phosphatidylserine (green) and liraglutide (red). The structure (as depicted) and mechanism of activation of GlpNP are hypothetical and speculated based on GlpNP physico-chemical characterization (see Supplementary Methods), the data below and the data presented in the main text. B) In a cell-free model system, biotin-containing non-fluorescent GlpNP were immobilized on 10 µm avidin latex beads, followed by annexin V-FITC staining in the presence or absence of gelatinases (cocktail of MMP-2 and MMP-9). GlpNP without gelatin coating showed uniform annexin V staining while gelatin coated GlpNP decorated with PtdSer demonstrated punctate annexin V staining following MMP-2/MMP-9 incubation, indicating degradation of the gelatin and binding of annexin V to the exposed PtdSer Fluorescence micrographs showed prominent punctate Annexin V-FITC staining of GlpNP decorated with phosphatidylserine (PtdSer+) and gelatinase-degraded gelatin (Gelatin + MMP) [far right]. Nanoparticles without gelatin coating but with PtdSer+ showed uniform staining [middle]. No staining was observed in the absence of PtdSer [left]. C) GlpNP that contained both PtdSer and degraded gelatin coating were engulfed rapidly. Low uptake was noted with an intact gelatin layer (PtdSer is not exposed) in the absence of MMP-2/MMP-9 D) Macrophages engulfed GlpNP that contained both PtdSer and degraded gelatin coating, but not GlpNP where gelatin was intact. Uptake efficiency was quantified by imaging flow-cytometry after treatment of bone marrow-derived macrophages with nanoparticles in the presence or absence of MMP2/MMP9.      Gelatin neg GlpNP were administered at 30 µg/kg (based on liraglutide) via bolus i.v. injection (left panel) and the mice were euthanized 24 h later. Gravity perfusion for at least 10 min with citrate-PBS was used to clear any non-engulfed particles before organ collection. In a separate experiment, 300 µg/kg of free liraglutide was administered simultaneously with 30 µg/kg Gelatin neg GlpNP (right panel) and the organs were imaged as above.

Figure S15. Immunofluorescence analysis of nanoparticle cellular distribution in atherosclerosis. A,B)
GlpNP localization in BCA 24 h after bolus injection in high-fat fed Apoe -/mice as described in the text and methods. Frozen aortic root sections were stained with antibodies against LGALS3 (Mac 2, Galectin-3, cyan) and αSMA (red) and co-visualized with endogenous Cy5 signal from GlpNP (green). The pattern of GlpNP distribution (arrows) was within αSMA+ fibrous cap, which was also enriched in macrophage-like LGALS3+ cells. Nuclei (blue) were stained with DAPI.

Materials, supplies and general methods
Extensively validated monoclonal anti-human GLP1R antibody (MAb 3F52) [1] was a kind gift from Novo Nordisk, Denmark. Lipoproteins LDL, oxLDL, and HDL were from Athens Research and Technology (Athens, GA). Liraglutide was from Bachem Americas (Torrance, CA). PCR reagents and cell culture supplies were from Life Technologies as indicated (see Supplementary  Table S1). Mouse antibodies (Supplementary Table S1) were from Abcam (Cambridge, MA), Thermo Fisher, and BD Biosciences. Bone marrow and peritoneal macrophages (BMDMs and P-Mac respectively) were obtained according to published protocols and as previously described by us [2][3][4]. HPLC solvents were LCMS Optima grade and were from Thermo Fisher. Insulin concentration was analyzed using ultra-sensitive ELISA (Crystalchem, Elk Grove Village, IL). Cyclic AMP was measured by an ELISA kit from Cayman Chemical (Ann Arbor, MI). Mouse Apolipoprotein B48 ELISA Kit was from MyBioSource, Inc. (San Diego, CA). Laser capture microdissection was performed by The Laser Capture Molecular (LCM) Core of The Ohio State University Medical Center (OSUMC). Myograph experiments were performed on thoracic aortas as previously described by us. [5] Tissue processing, staining and immunohistochemistry was performed by the Case Western Reserve University Institute of Pathology core laboratory. Magnetic resonance imaging of the abdominal aorta was performed as previously described by us [4,6] and others [7][8][9] using Bruker BioSpec 9.4T horizontal bore scanner.

Immunohistochemistry of human plaque
Human tissues were obtained from autopsy specimens, under exemptions granted by the Institutional Review Boards of the New York-Presbyterian Weill Cornell Medical Center. Immunohistochemical staining was performed using GLP1R (MAb 3F52, Novo Nordisk, Denmark) staining in human coronaries that were formalin-fixed and paraffin-embedded. Staining was performed on the 5 micron tissue sections on a Leica Bond system using the standard protocol and as described by us. [10] The sections were pre-treated using heat mediated antigen retrieval with a Tris-EDTA buffer at pH 9 for 20 min. The sections were then incubated with MAb 3F52 (1:1000 dilution) for 15 min at room temperature and detected using an HRP conjugated compact polymer system. 3, 3-Diaminobenzidine (DAB) was used as the chromogen. The sections were then counterstained with haematoxylin and mounted with Leica Micromount. Alternatively, immunohistochemistry was performed on the Ventana Discovery system (Roche). Primary anti-human GLP1R mouse antibody (Mab 3F52, Developmental Studies Hybridoma Bank) was used at 1:5 dilution. The staining was developed and visualized by using the NovaRED kit (Vector Laboratories).

Immunofluorescence in human plaque
Formalin fixed paraffin-embedded coronary artery tissues were selected from the CVPath autopsy registry. Histologic sections were cut on a rotary microtome at 5 to 6 μm, mounted on charged slides, and stained with hematoxylin and eosin (H&E) for tissue characterization.
Primary anti-human αSMA rabbit antibody obtained from Abcam (ab5694) was used with 1:1000 dilution. Primary anti-human GLP1R mouse antibody (Mab 3F52, Developmental Studies Hybridoma Bank) was used in 1:5 dilutions. Alexa Fluor-labeled secondary antibodies (Life Technologies) were used for further incubation. Tissue was visualized using an objective on a LSM 800 laser scanning confocal microscopy (Zeiss). Projection images were generated by collecting the maximum pixel intensity from each image of the Z stack and by projecting pixel intensity onto the single (projection) image.

Human peripheral blood monocytes (PBMC) isolation and differentiation
PBMC isolation and culture was performed as previously described. [11,12] Briefly, blood was drawn from healthy volunteers into BD Vacutainer blood collection tubes with sodium heparin. Mononuclear cells were then isolated using a sterile density gradient centrifugation with Ficoll-Paque PLUS (density 1.077 g/mL) according to the manufacturer's instructions. PBMC cell suspension was adjusted to a 2 x 10 6 cells/mL in RPMI 1640 containing 20% autologous serum. The cell suspension was transferred into 350 mL perfluoroalkoxy alkane (PFA) jars (Savillex Corporation, Eden Prairie, MN) and closed loosely with screw caps. The jars were placed into a 5% CO 2 incubator at 37 o C for five days, during which PBMCs differentiated into macrophages. Next, the cells were washed with cold RPMI 1640 three times via centrifugation at 300 g for 10 minutes and 4 o C. The cells were counted and seeded onto plates at indicated densities using RPMI 1640 with 10-20% autologous serum and 50 ng/mL human macrophage colony stimulating factor (M-CSF, Miltenyi Biotec) to allow for complete differentiation. The experiments were performed on day seven of differentiation.

Primary mouse carotid artery endothelial cells
Primary mouse carotid artery endothelial cells were purchased from Cell Biologics Inc. (Chicago, IL) and cultured using complete mouse endothelial cell medium (Cell Biologics Inc.) on gelatin-coated dishes and plates and according to the vendor's recommendations.

Quantitative real-time PCR
Total RNA was extracted from cells using the Absolutely RNA Miniprep Kit (Agilent-Stratagene) according to the manufacturer's protocol, and treated with DNase I to remove residual DNA. Total RNA from tissue was extracted using TRIzol (Thermo Fisher). RNA from magnetically-separated cells was extracted using the Absolutely RNA nanoprep kit (Agilent Technologies). Total RNA (>0.5 µg) was reverse-transcribed and cDNA was amplified by PCR using the High Capacity cDNA Reverse Transcription Kit from Thermo Fisher. The sequences of the primers and TaqMan probes are listed in Supplementary Table S1. Gene expression was quantified using the Roche LightCycler 480 detection system. Amplification was detected with SYBR Green as a fluorogenic probe specific for double-stranded DNA by using a Power SYBR Green PCR Master Mix (Roche Diagnostics GmbH). Gene amplification was normalized to 2-3 different housekeeping genes and the best stable control genes were identified using geNorm. [13] Relative copy number (RCN) and fold change were determined. RCN was calculated as follows: RCN = E-ΔCt × 100, where E is the efficiency (E = 10^[-1/slope] 2.0 = 100% efficient) and ΔCt = Ct(target) -Ct(reference). Duplicate samples were analyzed in each experiment.

Oral glucose tolerance test (OGTT) and glycemia metrics
Mice were fasted with full access to water for 16 hours before oral gavage with glucose at 2 g per kilogram body weight. Mice were not anesthetized during gavage. Nanoparticles or liraglutide were injected 2 hours prior to OGTT. Fasting blood glucose was measured using a hand-held glucometer (Ascensia Diabetes Care, Parsippany, NJ) from saphenous vein blood collected before gavage and re-sampled every 5-10 minutes for two hours after the gavage. Approximately 20 µL of blood was collected at every time point into an EDTA capillary tubes for the analysis of blood insulin concentrations using ultra-sensitive enzyme-linked immunosorbent assay (ELISA, Crystal Chem).

Flow cytometry, fluorescence-activated cell sorting (FACS)
Single-cell suspensions from an aortic sample or whole blood were prepared as previously described. [4] Briefly, perfused aortas were cut lengthwise and finely minced in <2 mm pieces in a well of a 6-well plate. The pre-warmed enzyme cocktail consisting of liberase (4 U/mL), DNAse I (50 U/mL) and hyaluronidase (60 U/mL) in 2 mL of serum-free RPMI was then added and incubated for 1 h at 37 o C. The pieces of semi-digested tissue were then agitated with pipetting using a large-orifice pipette tip and additionally incubated for 30 min. The mixture was then added to a 10 mL of cold RPMI containing 5% FBS and centrifuged (200 g x 10 min, 4 o C). The cell pellets were resuspended in RPMI+ 5% FBS and passed through a 50 µm cell strainer before repeating the centrifugation step. Cell suspensions containing 10 6 cells were put in 1.5 ml centrifuge tubes and diluted to 1 ml. The tubes were centrifuged (400 g, 5 min, 4 o C) and all but 100 μl of supernatant was carefully removed. The cells were resuspended by gentle tapping and then 1-3 μl of each FACS antibody was added. Cells were incubated for 20 min at 4 o C with gentle tapping every 5 min. Next, 1 mL of FACS buffer (1% FBS in DPBS) was added and tubes were centrifuged (400 g, 5 min, 4 o C). The cells were washed twice by resuspending in the FACS buffer and centrifuging. Finally, cells were fixed in 0. 5

Lipid analysis
Plasma lipids were analyzed using fast protein liquid chromatography (FPLC). The FPLC system consisted of the Gilson GX271 Liquid Handler (Gilson Inc, Middleton, WI) equipped with a PHD ULTRA HPSI/XF high precision, high pressure syringe pump (Harvard Apparatus, Holliston, MA). Separation of lipoproteins was performed on a Superose 6 Increase 3.2/300 column (GE Healthcare Bio-Sciences, Pittsburgh, PA). Plasma samples were diluted ten times with a phosphate buffered saline followed by an injection into the column using a Gilson GX Direct Injection Module equipped with a 10 µL injection loop. The reproducibility of the separation was achieved using "loop volume overfill" task, which overfilled the loop with a 40 µL of the sample before each injection. Lipoproteins were eluted in an isocratic mode with a buffer containing 10 mM phosphate and 154 mM NaCl at pH 7.4 and room temperature. The flow rate was 0.05 mL/min. Fifty 100 µL fractions were collected in opaque 96-well plates positioned on a cooling block at 4°C. The fractions were then analyzed for cholesterol and triglyceride concentrations using hydrogen peroxide detection assays based on the peroxidase-oxidase reaction as previously described [14,15] and briefly outlined below.
Plasma samples or FPLC fractions were incubated with cholesterol or triglyceride reagents that were prepared from fresh lyophilized enzymes on the day of analysis as follows (see Supplementary Table S1 for enzyme details). The cholesterol reagent consisted of 25 mM PIPES at pH 6.7, 250 U/mL cholesterol oxidase, 1000 U/mL cholesterol esterase, 1 U/mL peroxidase (horseradish), 5 mM sodium taurocholate, and 0.25 mg/mL Amplex Red (10-Acetyl-3,7-dihydroxyphenoxazine). The triglyceride reagent was prepared in 25 mM PIPES at pH 6.7 with 2.5 mM ATP, 2.5 mM magnesium acetate, 10 U/mL glycerol 3-phosphate oxidase, 5 U/mL glycerokinase, 100 U/mL lipase, 1 U/mL peroxidase (horseradish), and 0.25 mg/mL Amplex Red. Fifty microliters of the sample (FPLC fraction or 10x diluted plasma sample) was incubated with 150 µL of either cholesterol or triglyceride reagent in opaque 96-well plate along with the appropriate clinical-grade standards at a range of concentrations of 1 -1000 mg/dL. The standards were from Stanbio Laboratory (Boerne, TX) or Pointe Scientific (Canton, MI) as indicated in the Supplementary Table S1. The plate was incubated at 37°C for 30 min and the Amplex Red fluorescence was recorded at 544 nm excitation and 594 nm emission using SpectraMax i3 plate reader (Molecular Devices, San Jose, CA). The concentration of cholesterol or triglycerides was derived from the respective standard curves using SoftMax Pro 6 software and plotted against FPLC fraction number. The total concentration of lipids in vLDL, LDL and HDL fractions were determined by the integration of the FPLC peaks and obtaining the area under the curve.

Liraglutide and GlpNP proteolytic stability assays
Liraglutide or GlpNP were resuspended at 0.02 mg/mL (based on liraglutide concentration) in an incubation buffer consisting of 20 mM HEPES buffered saline pH 7.4, 1 mM magnesium chloride, 1.5 mM zinc chloride and 1% pooled human plasma (Red Cross). 90 µL of each solution was transferred to a vial equipped with a magnetic stir bar and containing 10 µL of 0.4 mg/mL CHO-expressed Neprilysin (Bio-Techne Corporation, Minneapolis, MN). In some experiments, recombinant MMP2 (1 µg/mL, Bio-Techne) plus MMP9 (1 µg/mL, Bio-Techne) were used. The solution was stirred at 37 o C and 5 µL aliquots were taken at predetermined time points (See Figure 2B). The aliquots were then quenched with 45 µL of 1% trifluoroacetic acid (TFA) in 80% acetonitrile / 20% water. The quenched solutions were transferred in 0.45 µm PTFE mini prep filter vials (Analytical Sales and Services, Flanders, NJ) and filtered. Next, the concentration of liraglutide was determined using high performance liquid chromatography (HPLC) as described below. In gelatinase degradation experiments, aliquots were not taken, but rather 10 µL of GlpNP particles were diluted in 0.5 mL PBS and spun down at 100,000 g (10 min, 4 o C). The supernatant was aspirated by vacuum suction and the nanoparticle pellet was processed in TFA/acetonitrile/water as above.
The separation was performed using the Shimadzu Prominence system consisting of three LC-20AD pumps, autosampler, RF-20Axs fluorescence detector and Sedex 90LT evaporative light scattering detector (ELSD). The sensitive detection (detection limit >2 pmol of liraglutide) was possible due to intrinsic fluorescence of liraglutide due to the presence of tryptophan in the peptide sequence, which was detected at 280 nm excitation and 350 nm emission wavelengths respectively. Five microliters were injected into 4.6 mm x 50 mm Promix MP column (SIELC Technologies, Wheeling, IL), which was equilibrated and maintained at 40 o C using a column heater. The mobile phase was delivered at 1 mL/min using three pumps: A 100% water, B 100% acetonitrile and C 1% TFA. The gradient was formed by pumping from 15% B to 45% B over 25 min, while concentration of TFA was maintained constant at 10% C (0.1% of TFA total). The RF-20Axs peaks were manually integrated using Shimadzu LabSolutions software and the data were expressed as a percent of the liraglutide peak area at time 0 (initial concentration). The peaks from the ELSD detector were not used for quantification and served as a qualitative assessment of the formation of low molecular weight degradation products that are not possible to register with the UV or fluorescence detector.

Synthesis of acetylated LDL and BODIPY-cholesterol-acetylated LDL
Acetylated LDL was prepared as previously described. [17] Briefly, 1 mL of 10 mg/mL of LDL was combined with 1 mL saturated sodium acetate solution and the mixture was stirred on ice at 200 rpm using a magnetic stir bar. Acetic anhydride was then added in 2 µL portions and 10 min intervals to a final concentration of 6 µL acetic anhydride per 1 mg LDL protein. The acetylated lipoprotein solution was dialyzed against 10 mM phosphate buffer plus 154 mM NaCl. Next, 50 µL of 50 mg/mL BODIPY cholesterol (TopFluor cholesterol, Avanti Polar Lipids) in 2-propanol was added to 1 mL lipoprotein deficient serum and stirred at 37°C for 4 h. Five milliliters of acetylated LDL at 1 mg/mL was then added to the above mixture and incubated at 37°C overnight with magnetic stirring. BODIPY-cholesterol-acetylated LDL was then isolated by ultracentrifugal flotation (4 h, 400,000 g at 4°C) with the density adjusted to 1.063 g/L. The centrifugation was performed using Thermo Scientific Sorvall MX 120 Plus ultracentrifuge and S120-AT3 rotor. The product was collected as a top band visible in UV light. The final product was dialyzed overnight against 10 mM phosphate buffered saline at 4°C and filtered through a 0.45 µm cellulose acetate membrane using a syringe filter before treating the cells  -100-042). The cells were seeded onto opaque 96-well plates at 2 × 10 5 cells per well and cultured in 100 µL of medium at 37°C in a 5% CO 2 atmosphere for 24 h. Next, HASMC were cholesterol-loaded by incubation with 100 μg/mL BODIPY-cholesterol-acetylated LDL (obtained as described above) in the serum-free medium containing 1.5% human serum albumin for 48 hours. One day prior to the cholesterol efflux experiments, the cells were incubated for 24 hours in the medium containing 1.5% human serum albumin and in the presence or absence of 10 nM liraglutide, GlpNP (containing 10 nM liraglutide and 0.1 mg/mL polystyrene nanoparticles), CtrlNP (0.1 mg/mL polystyrene nanoparticles), and medium alone. In some experiments, 1 hour prior to incubation with nanoparticles, 1 µM cAMP inhibitor (3-Isobutyl-1-methylxanthine) or 100 nM GLP-1R antagonist (exendin fragment 9-39) or 10 µM forskolin was added. Next, the medium was removed and replaced with 100 µL of fresh culture medium containing only Vascular Cell Basal Medium and 1.5% human serum albumin. This was repeated twice over the next 1 hour. Then, these BODIPY-cholesterol-acetylated LDL-loaded HASMCs were incubated in 100 μL serum-free medium with or without apoA-I (50 μg/mL) or HDL (100 μg/mL) for 24 hours. The fluorescence at excitation of 490 nm and emission of 515 nm of the plates was recorded. The result of this reading is Total Cell Fluorescence (TCFl). The efflux medium was then transferred to a new opaque 96-well plate (or to the empty wells of the same plate) with a multichannel pipette. Fluorescence at excitation of 490 nm and emission of 515 nm was read, resulting in Efflux Medium Fluorescence (EMFl

RNA sequencing
For bulk RNA-seq, samples were lysed with Trizol reagent and the RNA was extracted and purified as described above. Full length RNA-seq libraries were prepared by Novogene corporation and sequenced using Illumina HiSeq platform. Raw data were recorded in an FASTQ file and analyzed as previously described. [18] The expression information on specific genes of interest was extracted from normalized data (RPKM) using R version 3.6.0 and dplyr package and plotted against normalized densitometry data from Southern blotting experiments (see above). All data associated with this study have been deposited in the Sequence Read Archive (SRA) at NCBI under the BioProject IDs: PRJNA544396 and PRJNA678544.

Atherosclerosis studies and analysis
Morphological analysis of atherosclerotic plaque and immunohistochemistry was performed as previously described. [19] The stained slides were imaged using Keyence BZ-X700 microscope and analyzed using Keyence BZ-X Analyzer software ver. BZ-H3XD. Hybrid cell count mode was used to determine areas of plaque, necrotic core or F4/80 (or αSMA)-positive cells. Next, the selection was made using Extraction by Brightness mode and the threshold was set at 864. Unwanted areas were removed by filtering by area and shape. Modes such as "fill cracks" and "deburring" were further used to refine the selection. BZ-X Analyzer output was area in µm 2 or mm 2 and number of total areas. The data were analyzed using repeated measures ANOVA with Shapiro-Wilk test of normality and 3-way between-subjects ANOVA using fixed-effects model. Equality of variances was tested using the Bonferroni-Dunn post hoc test.

Myography
Myographic analysis of the isolated aortic rings was performed as previously described. [20,21] Briefly, mice were anesthetized and the thoracic aorta was removed, cleaned and placed in physiological salt solution (PSS) containing: 130 mM NaCl, 14.9 mM NaHCO 3 , 4.7 mM KCl, 1.18 mM KH 2 PO4, 1.18 mM MgSO 4 , 1.56 mM CaCl 2 , 0.026 mM EDTA, and 5.5 mM glucose. The tissue was cut into 2-mm rings that were mounted in a chamber of a Multi Myography System (Danish Myo Technology A/S) containing PSS at 37°C and continuously aerated with 95% O 2 -5% CO 2 . Isometric force generation was recorded with a resting tension of 30 mN, and the rings were allowed to equilibrate for 2 h. Arterial integrity was assessed first by stimulation of vessels with 80 mM KCl. Endothelium-integrity was assessed by measuring the dilatory response to 10 μM acetylcholine in vessels pre-contracted with 1 μM phenylephrine. To test aortic relaxation, the rings were pre-contracted with 0.3 μM phenylephrine that induced approximately 60% of maximal contraction, which was followed by addition of acetylcholine (ACh) in a cumulative manner and at the following doses in nM: 1, 3, 10, 30, 100, 300, 1000.

Cholesterol analysis in atherosclerotic plaque
Free cholesterol concentration in atherosclerotic plaque was obtained after extraction, derivatization with pentafluorophenyl isocyanate (PFPI) and analysis by liquid chromatography/mass spectrometry (LCMS) as previously described. [19,22,23] Briefly, excised atherosclerotic plaques were homogenized in 1 ml of hexane-isopropanol 2:1 (v/v) with 1 nM cholesterol-d 7 added as an internal standard. The extraction was repeated with obtained precipitates two more times and combined extracts were centrifuged, filtered and evaporated under a stream of argon gas. The residue was re-dissolved in 1 mL 100% hexane and 3 µL pyridine and 5 µL of PFPI was added. After 10 min at room temperature, 100 µL of 100% isopropanol was added to quench the reaction and dissolve the precipitate.
Five microliters of derivatized extracts were injected into Agilent Zorbax Rx-SIL column (150 x 4.6 mm, 5 µm) and separated using Shimadzu Prominence HPLC (see above) using an isocratic solvent system consisting of 90/10 hexane/isopropanol and a flow rate of 1 mL/min. The effluent was analyzed on the AB Sciex 3200 QTRAP LC/MS/MS system coupled with an atmosphere pressure chemical ionization (APCI) ionization mode. Mass spectra were acquired in negative ion mode using a capillary temperature set at 300 o C, and spray voltage was set at 3.5 kV. Nebulizer gas and auxiliary gas flow were set to 60 and 20 arbitrary units respectively. Curtain gas (CUR) was set to 15.00 and collision gas (CAD) settings were set to medium. Detection was by multiple reaction monitoring (MRM) with the transitions monitored 595.4→206.0, 602.5→206.0 corresponding to cholesterol and cholesterol-d 7 respectively. Total cholesterol was analyzed after hydrolysis of cholesteryl esters with sodium methoxide as previously described, [22] levels were analyzed as described above.

Statistical analysis
We examined an N of 6-12 per group for cell-based in vitro assays and indicated in figures N for each group for mouse experiments. Data were tested for normality using the D'Agostino-Pearson normality test and for equality of variances using Bartlett's test. If normality and equality of variances are not rejected at the significance level of 0.05, the group means were compared using Student's t-test (for 2 groups) or pairwise t-test (for >2 groups) with Holm post hoc correction or ANOVA with Tukey's test (for >2 groups). For non-normally distributed data or data with unequal variances, nonparametric Mann-Whitney U-test (for 2 groups) and Kruskal-Wallis test and Dunn's test (for >2 groups) all with Holm post hoc correction.  Supplementary Table S2: Screening of different nanoparticle formulations by varying concentrations of gelatin, PEG, and the core (polystyrene nanoparticles). The criteria for the "best" formulation were a) the smallest hydrodynamic diameter, b) the l owest polydispersity, and c) highest liraglutide i ncorporation. Five "lead" formulations are shown out of 32 different formulations screened (data not shown). Highlighted i n yellow i s the GlpNP used i n this paper.  Supplementary Table S3: Results of the correlation analysis between Glp1r expression by Southern blotting (see Figure 1M of main manuscript) and the l evels of various myeloid and smooth muscle-related gene mRNA transcripts (as i ndicated) determined by RNAseq. Correlation values are shown as R^2 and p values were determined from multiple comparisons of means (ANOVA with Tukey's post-hoc test).